The people demand random bits of knowledge and I accomodate. I'm kicking around some ideas about the mechanism of memory storage in the brain, and if you haven't read the background post about Memory Consolidation you can do that first.
So, here's a few things you should know beforehand. It's possible to stick electrodes in a rat's brain and record the activity of single cells as the animal walks around and does what he does (you can distinguish more than 100 cells at a time if you're really good - the different colored clusters on the figure at left are responses from different cells). And it doesn't hurt the rat, he loves it, actually - they give him chocolate. Anyway, if you do this and stick electrodes in the animal's hippocampus, you can find cells that respond only when the animal is in a particular location - whether he's in a maze or wherever. And you can record from many cells and find cells that respond when he's in one place, and some that respond when he's in another place, and some when he's in a third place, and some when he's in a fourth place, and so on... They call these place cells. (The figure at right shows the responses of different cells as the animal moves from one area to the next - different cells are active when he's in different places.)
So, people have done this (Lee and Wilson), and recorded where the animal goes and which cells fire. And then they record the activity of these cells while the animal is sleeping (during slow wave sleep i.e. deep sleep, with no dreaming). During slow wave sleep, there are slow oscillations in activity of most of the cells in the area, accompanied by sharp waves, where you find many cells firing at the same time. (Those fast downward spikes on the EEG pictures during the still period are the sharp waves.) And if you look at the same cells that you were recording from when the animal was moving around, you find they are participating in the sharp wave, and they fire in the same order they fired in when the animal was navigating his maze. The sharp wave is thought to originate directly in the hippocampus, so it seems like the hippocampus has recorded this activity, during the day, and plays it back during the night - over and over again, for long term storage somewhere else - presumably in the cerebral cortex.
There is one key difference between the activity of the place cells during the awake-behaving period to the activity during the sharp waves during sleep that should be mentioned. During awake navigation, the cells fire in order dictated by their place field in the time it takes him to move from one field to another - in this case the whole navigation of the 'maze' took about 4-5 seconds, so recording the whole sequence of place cell firing takes 5 seconds. But during sleep, during the sharp wave, they all fire in the same order, but it takes less than 200 milliseconds from the first to the last. So playback of this information was about 20X faster than recording - on average.
Interestingly, this speeding up of playback might just make it easier to understand how it gets stored in cortex, because neurons do everything fast. Action potentials, and biochemical pathways, everything takes milliseconds to reach completion. If an animal is going to learn that there is a relationship between two events (i.e. bar press = food reward), it's difficult because these things happen at a very slow timescale. And if it takes 200 milliseconds for the whole chain of events to be played back, it's a more reasonable task. And it ties the reasearch into a huge body of literature by people who study learning and memory at the single cell level - who find that co-activation of two cells within a time of 20 ms or so will lead to a strengthening of the connection between them.
So, basically, the idea is that, the hippocampus records a slow sequence of events and plays it back real fast, and the sequential firing of cells, drives cells in cortex to fire at the same time, and thereby strengthening the connections between the cortical cells. Thereby consolidating those memories there for long term storage.
Of course, saying that the hippocampus guides the long term storage of information in cerebral cortex begs the question: How does it get recorded in the hippocampus in the first place? But that's a whole different thing.
There's more to the story, that I'll get to tomorrow. There's some new data that came out last month that might throw a wrench in this model, or maybe not, it's unclear, it's cutting edge, you'll hear it here first.